What happens when the AI tool that wrote your product descriptions gets something wrong, and Google is the one who notices? That's no longer a hypothetical. This week Google tightened its guidance on AI-generated content, and it lands right on top of the shortcuts a lot of online shops have been taking. There's also news on loyalty programmes in AI Mode, cheaper ways into YouTube advertising, better testing in Microsoft Advertising and a sensible reminder about landing pages. Let's get into it.
Google Wants AI Content Fact-Checked, and Gemini Is Leaving Fingerprints
Google has updated its guidance to say that AI-generated content should be fact-checked by a person before it's published. The guidance covers more than blog posts and product descriptions. Google specifically names titles, meta descriptions, structured data and image alt text.
Why does that matter to you? Because those are exactly the places where many shops have handed the job to AI in bulk. Hundreds of product titles get generated in an afternoon, a plugin writes the alt text, and nobody has checked the structured data since it was installed. If any of that says a jacket is waterproof when it isn't, or lists the wrong size or material, that's a problem for Google. It's an even bigger problem for the customer who receives something different from what they expected.
My view is plain and simple. AI is a brilliant first-draft machine, but someone who actually knows your products needs to sign off the final version.
The same round-up flagged something else. People have spotted UTM tags (the little tracking codes added to the end of a link) on clicks from Google's Gemini to websites. Google's John Mueller acknowledged seeing them, but Google hasn't documented when or how Gemini adds them. If Gemini starts showing up in your analytics, that's probably why. It's also an early sign that AI assistants are becoming a traffic source you can actually measure, and that's worth keeping tabs on.
Your Loyalty Programme Can Now Follow Customers Into AI Mode
Google has updated its Merchant Center help documentation for loyalty programmes, and the biggest change is that Loyalty Customer Match now covers AI Mode and Gemini. In plain terms, Google can now show your loyalty members their member prices and perks in its AI-powered shopping experiences, as well as in standard search and Shopping results.
The programme has also expanded to more countries, including Italy, Spain, the Netherlands, India and South Korea, alongside markets like the UK that were already covered. Google has also made it clear that you don't need to be an advertiser to take part. Programmes need to be genuinely free to join, so a free trial that rolls into a paid subscription won't qualify.
What does this mean for your shop? If you run a VIP or points scheme, it's now a stronger reason to connect it to Merchant Center. As more product research happens inside AI Mode, a returning customer who sees "your member price" right in the answer gets a strong nudge to buy from you. If you don't have a loyalty scheme, a simple one is now worth more than it was. Repeat customers are almost always the cheapest ones to win.
Your Social Content Could Be Your Next YouTube Ad
In its latest Ads Decoded episode, Google talks about how creative can make or break a YouTube campaign. The main message is reassuring. Your existing social media creative is probably a solid starting point, as long as it meets Google's asset specs and guidelines.
Google also highlighted Asset Studio, its video creation tool powered by Gemini Omni. It can turn your website URLs and product photos into videos that you can preview before anything goes live. Google's advice is to avoid flooding campaigns with dozens of assets. Make changes gradually, put quality ahead of quantity and give the system time to measure what's working.
Many e-commerce owners stay away from YouTube because they assume it needs a production budget. It usually doesn't. You probably already have short product clips, unboxing videos or customer content that could be adapted. I agree with the "fewer, better" advice. If you change everything at once, you'll never know what actually made the difference.
Microsoft Advertising Makes Proper Testing Easier
Microsoft's September round-up has a few updates. A HubSpot integration and bigger LinkedIn company lists (up from 1,000 to 10,000 companies per list) will mainly interest B2B advertisers. The update that matters for online shops is that experiments are now fully available across Search, Shopping, Audience and Performance Max campaigns.
That means you can test changes to bidding, targeting or creative on part of your traffic and compare the results side by side on a single results page. You don't have to change everything and hope for the best.
Microsoft Advertising (the ads that appear on Bing and across Microsoft's network) often gets treated as an afterthought. But clicks there can be cheaper, and running a fair test on a Shopping or Performance Max campaign is a low-risk way to find out what works before you put more budget in.
Landing Page "Best Practices" Are a Starting Point, Not a Rulebook
PPC Hero makes a point I wish more people heard. Standard landing page advice is a good baseline when you have no data, when a page has obvious usability problems or when visitors can't quickly understand what you're offering. That advice includes matching your page to your ad, cutting distractions and putting the important information where people see it without scrolling.
The article also says when to question that advice. Be sceptical when you don't have enough traffic to test properly, when the conversions in your ad platform don't match your actual sales, or when customer feedback tells a different story from the numbers. Its suggested approach is to start with best practice, look at your own data, form a clear idea of what to test, test one thing at a time, fit it around your budget and seasonality, and write down what you learn.
For an online shop, your product and category pages are your landing pages. Don't redesign them because a blog post told you to. And with the busiest weeks of the year approaching, the point about seasonality really matters. Peak trading is not the time to start a big experiment.
What I'm Watching
One idea links all of this. AI is doing more of the heavy lifting, but human judgement matters more than ever. Google wants a person checking what AI writes. AI Mode is turning loyalty schemes into a competitive edge. Creative tools can turn existing content into video ads, but they still need a sensible testing plan behind them.
With Black Friday and Christmas getting closer, I'd do three things now. Check any AI-generated product titles, descriptions and structured data for accuracy. Connect your loyalty scheme to Merchant Center if you have one. And run any tests you want answers to before peak trading, not during it.
I'll also be watching Gemini as a traffic source. If those UTM tags become a regular feature, you'll finally be able to see how much AI assistants are actually sending your way, and that will change how everyone thinks about where their sales come from.